Wetter subtropics in a warmer world: Contrasting past and future hydrological cycles Natalie J. Burlsa,b,1 and Alexey V. Fedorovb aDepartment of Atmospheric, Oceanic, & Earth Sciences, Center for Ocean-Land-Atmosphere Studies, George Mason University, Fairfax, VA 22030; and bDepartment of Geology and Geophysics, Yale University, New Haven, CT 06511 Edited by Jeffrey P. Severinghaus, Scripps Institution of Oceanography, La Jolla, CA, and approved October 11, 2017 (received for review February 28, 2017) During the warm Miocene and Pliocene Epochs, vast subtropical strength, i.e., v remains approximately constant, the first term on regions had enough precipitation to support rich vegetation and the right-hand side of Eq. 2 tells us to expect that regions on Earth fauna. Only with global cooling and the onset of glacial cycles some currently experiencing moisture convergence and hence more 3 Mya, toward the end of the Pliocene, did the broad patterns of arid precipitation than evaporation will become even wetter, while re- and semiarid subtropical regions become fully developed. However, gions of moisture divergence and net evaporation (e.g., ocean current projections of future global warming caused by CO2 rise subtropical regions) will become even drier. This is the essence of generally suggest the intensification of dry conditions over these the familiar paradigm “Wet gets wetter, dry gets dryer” (1). For the subtropical regions, rather than the return to a wetter state. What subtropics, this thermodynamic scaling, in its simplest form, has makes future projections different from these past warm climates? different implications over land than ocean, e.g., see ref. 2, since for Here, we investigate this question by comparing a typical quadru- land regions, there is the constraint that mean P − E values must be pling-of-CO2 experiment with a simulation driven by sea-surface positive or near zero. That said, future climate projections generally temperatures closely resembling available reconstructions for the predict the maintenance or intensification of dry conditions over early Pliocene. Based on these two experiments and a suite of other most subtropical land regions under future global warming (3, 4). perturbed climate simulations, we argue that this puzzle is explained Accounting for spatial gradients in surface temperature and by weaker atmospheric circulation in response to the different ocean humidity changes brings the thermodynamic scaling, based on the surface temperature patterns of the Pliocene, specifically reduced first term in Eq. 2, more closely in line with simulations of both meridional and zonal temperature gradients. Thus, our results high- future warmth (2) and cold conditions of the Last Glacial Maxi- light that accurately predicting the response of the hydrological cycle mum (5). Dynamical P − E changes due to the second term in Eq. 2 to global warming requires predicting not only how global mean are generally regarded as secondary. While climate theory, based temperature responds to elevated CO2 forcing (climate sensitivity) on global mean thermodynamic constraints, predicts that we should but also accurately quantifying how meridional sea-surface temper- expect some weakening of large-scale atmospheric overturning ature patterns will change (structural climate sensitivity). circulations in the tropics with global warming (1), in the zonal mean this thermodynamically explained weakening is generally not Pliocene | hydrological cycle | climate change | subtropical aridity strong enough to compensate for humidity increases and so the first term in Eq. 2 tends to dominate (3). Circulation changes do how- ollowing the Clausius–Clapeyron relation, Earth’s atmo- ever appear to play a role within regional subtropical P − E changes Fsphere can hold more water vapor in a warmer climate, past and the poleward expansion of subtropical aridity due to Hadley or future. Moreover, in steady state, the hydrological cycle at any cell widening in simulations of future climate (3, 6, 7). one point on the surface of the Earth must obey the following The drier subtropics scenario projected by climate simulations moisture budget, with the residual between precipitation (P) and of the future sits in contrast to evidence suggesting that the evaporation (E) balanced by the convergence of moisture in the overlying atmosphere: Significance Z 1 ps P − E = −∇ · vq dp, [1] The subsiding flow of the atmospheric Hadley circulation con- g 0 trols dry conditions over vast subtropical bands where the v main arid regions of the globe reside. In the context of future where g is gravity, q the atmospheric specific humidity, and the changes in the atmospheric hydrological cycle, understanding horizontal wind vector integrated across pressure (p) levels from precipitation changes in the subtropics is of particular impor- the bottom to the top of the troposphere (as shown schematically tance given the scarcity of water resources in these locations. A in Fig. 1). This budget tells us that in a warmer climate, changes major puzzle arises when contrasting the drier conditions in in net precipitation minus evaporation over a specific region, δð − Þ the subtropics predicted by climate models under global P E , will result from either changes in the water content warming scenarios against the wetter conditions seen in re- of the air circulating over the region or changes in the circulation constructions of past warm climates, including the warm, itself, or some combination of both: ∼400 ppm CO2, Pliocene. Our modeling results address this Z Z Z puzzle and highlight the importance of correctly predicting 1 ps 1 ps 1 ps δðP − EÞ = −∇ · vδq dp −∇· δvq dp −∇· δvδq dp, how the meridional temperature gradient between the tropics g 0 g 0 g 0 and the subtropics will change with global warming. [2] Author contributions: N.J.B. and A.V.F. designed research; N.J.B. performed research; and where δ represents the change in each variable between the N.J.B. and A.V.F. wrote the paper. warmer climate state and the reference climate state (which in The authors declare no conflict of interest. this study is a preindustrial climate). This article is a PNAS Direct Submission. According to Clausius–Clapeyron, and assuming more or less Published under the PNAS license. constant relative humidity as suggested by climate models, q is 1To whom correspondence should be addressed. Email: [email protected]. ∼ expected to increase by 7% per degree Celsius as climate This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. warms. If large-scale circulation patterns remain similar in 1073/pnas.1703421114/-/DCSupplemental. 12888–12893 | PNAS | December 5, 2017 | vol. 114 | no. 49 www.pnas.org/cgi/doi/10.1073/pnas.1703421114 Downloaded by guest on September 26, 2021 4xCO Pliocene A 2 B -δv -δv --δv --δv Increased Increased Decreased Decreased moisture moisture moisture moisture transport transport transport transport δvq < vδq δvq < vδq δvq > vδq δvq > vδq Fig. 1. Different response of the hydrological cycle P-E -δv ++δq -δv P-E P-E --δv ++δq --δv P-E to global warming due to differences in the meridi- 25°S Equator 25°N 25°S Equator 25°N onal structure of the warming. Based on (A and C) × Wet gets wetter, Dry gets drier Dry gets wetter, Wet gets drier the 4 CO2 and (B and D) Pliocene experiments. The schematics in the top panels summarize the net hy- 4xCO2 - Preindustrial Control SST Pliocene - Preindustrial Control SST C D °°° 90°°° E 180 E 270 E 90 E 180 E 270 E 907 drological cycle response to the SST anomalies (rel- 60° N 60° N 982 ative to the control) shown in the bottom panels. 883/884 1021 The reconstructed early Pliocene (4–5 Ma) SST 30 °N 30 °N 1014 607 1208 1010 958A anomalies from sites around the globe (SI Appendix, 1241 ° ° 722 850 925 Table S2) are superimposed on the Pliocene SST 0 0 806 846 1084 709 U1338 1237 change simulated by the coupled simulation with 214 590B 847 ° ° 1085 763A modified cloud albedo which is then used to force 30 S 30 S 1088 1125 594 the atmosphere-only Pliocene simulation. Surface 60° S 60° S temperature changes from the atmospheric model oC over both the land and ocean are shown in SI Ap- -1 012345678910 pendix, Fig. S1. hydrological cycle behaved in roughly the opposite manner Pliocene and Miocene reconstructions reflect a direct effect on during past warm periods. Wetter subtropics were a salient plants of changes in atmospheric CO2 concentration versus hy- feature of the warm Miocene, e.g., refs. 8–11—the subsequent drological cycle changes. Recent studies that do account for the aridification in these regions and the expansion of C4 grasslands physiological effects of reduced Last Glacial Maximum CO2 occurred as climate gradually cooled. Although subtropical ari- concentrations on pollen reconstructions of moisture availability dification appears to have begun in the Miocene and strength- have shown that better agreement is found with independent ened across the Pliocene and Pleistocene, we focus here on the paleohydrological reconstructions, e.g., ref. 26 and climate sim- more recent Pliocene Epoch, primarily because a sufficient ulations (27). That said, the enhancement of photosynthetic number of continuous sea-surface temperature (SST) recon- carbon uptake in C3 plants under increasing CO2 concentrations structions, dating back to the Pliocene from regions across the appears to be nonlinear with maximum sensitivity in the ∼0–250- global ocean, allow us to characterize the nature of large-scale ppm range (28). Therefore, with reconstructed Pliocene CO2 SST gradients (11–13). One of the central large-scale features of concentrations falling within the ∼350–450-ppm range (11, 12, Pliocene warmth is that it was polar amplified with meridional 19), neglecting the CO2 fertilization signal in pollen-based re- SST gradients much weaker than present, including those be- constructions is probably a relatively small source of error since, tween the tropics and the subtropics as the warm pool was ex- as summarized in SI Appendix, Table S1, subtropical recon- panded (13–15).